1
$\begingroup$

In handwritten digit recognition problem using logistic regression, normal implementation would forcibly classify even a picture of dog or cat as a digit. To eliminate this, what changes are needed to add another class i.e. "Not a Digit" to already existing 10 classes (0 to 9) ?

$\endgroup$
1
  • $\begingroup$ I personally would start by adding letters as "letters". $\endgroup$ – James Phillips Nov 16 '18 at 17:42
0
$\begingroup$

Such implementation would only require adding one more class for the target variable.

Further training data should have examples from the "Not A Digit" class as well(with appropriate label).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.